BoosTexter is a general purpose machine-learning program based on boosting for building a classifier from text and/or attribute-value data. BoosTexter can handle:

Note, however, that BoosTexter uses boosting on top of very simple decision rules (sometimes called "decision stumps"). Although this allows BoosTexter to run very fast while often giving highly accurate results, this approach may not be appropriate for all learning tasks. For instance, boosting on top of decision trees (such as C4.5 or CART) may be more effective for some applications.

For more information, contact Mazin Gilbert (mazin at research dot att dot com)

The object code for BoosTexter is available free for non-commercial research or educational purposes by clicking here.

BoosTexter was written by Erin Allwein, Robert Schapire, and Yoram Singer.